Merge pull request #1100 from deepseek-harness/codex/web-stop-preserve-queue

fix(web): preserve queued prompts when stopping
2026-07-31 21:51:25 +08:00
parent fb9da7587c 32972f1675
commit 880203a738
20 changed files with 241 additions and 50 deletions
@@ -0,0 +1,6 @@
+# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
+# side as of the last confirmed-consistent state. Both languages carry equal authority;
+# after editing either side, bring the other along and re-record with:
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/bug-fix/2026-07-31-web-stop-preserves-queue.md
+2026-07-31-web-stop-preserves-queue.md: 943e95d6951a28929c4f8ce4d0b6e17224b08ede
+2026-07-31-web-stop-preserves-queue.zh.md: bbadd8adf8fd5bb0604ef87d322e48ce4c2ed759
@@ -0,0 +1,37 @@
+# Agent Note: Web stop preserves pending Queue
+
+Status: implemented
+
+English | [中文](2026-07-31-web-stop-preserves-queue.zh.md)
+
+## Problem
+
+The Web stop button reached `session.cancel`, which mapped to broad `agent.cancel({ kind: 'user' })`. During an active turn, ordinary composer submissions are already accepted as independently addressable Queue occurrences. Broad cancellation discarded every occurrence when the user intended to stop only the current generation, conflating turn interruption with the Queue's explicit delete operation.
+
+The browser cannot repair that loss by resending visible rows. It does not own their live `InboxItemId`, wake policy, or claim race, and a resend can duplicate work that the Host has already claimed.
+
+## Decision
+
+`session.cancel` is the Web Host API's active-turn stop. It calls `agent.cancel({ kind: 'user' }, { keepInbox: true })`, preserving pending inbox work while cooperatively aborting the current turn. The underlying option preserves queued and steering entries; the Web Queue projection continues to expose only queued entries.
+
+The AgentLoop starts no concurrent replacement turn. It closes and flushes the interrupted turn, reaches cancellation quiescence, and then claims the next waking queued occurrence through its existing FIFO driver. That claim emits `agent/inbox/dequeue`, so the Host's authoritative `session/queue` snapshot retires the claimed row and leaves the remaining tail visible. The browser neither resends nor promotes any row. Work that ignores cancellation delays this handoff until it settles.
+
+This mapping changes only the Host `session.cancel` endpoint used by Web clients. The default `Agent.cancel()` contract remains broad, ACP and TUI retain their existing cancellation policies, and `AgentHandle.dispose()` still clears pending work during teardown. Queue row removal remains the explicit Web action for discarding one pending occurrence.
+
+## Alternatives considered
+
+**Keep broad cancellation for the stop button.** Rejected because stopping one generation should not destroy independently queued user intent; the Queue already owns explicit deletion.
+
+**Resend the next row from the browser after cancellation.** Rejected because the Host owns occurrence identity and claim order. Client resubmission can duplicate work, reorder the FIFO, or race an authoritative dequeue.
+
+**Start the next turn before cancelled work reaches quiescence.** Rejected because two turns would concurrently mutate one session log and share Agent-owned resources. Cooperative cancellation waits truthfully for the active work to settle.
+
+**Add a wire option for broad versus preserving cancellation.** Rejected until the Web product has a separate “stop and clear Queue” interaction. The existing stop button has one policy, while per-row delete already supplies the current discard control.
+
+## Verification
+
+AgentLoop coverage holds an active model stream, queues two waking turns, cancels with `keepInbox`, and pins the aborted-then-completed turn reasons, FIFO user-message order, absence of discard events, and eventual idle state. The keyless Web scenario drives the built composition over HTTP/SSE: it stops one hung turn, observes the next queued occurrence start while the tail remains visible, stops that turn, and observes the final queued occurrence complete. Its accessibility snapshot pins the intermediate preserved-Queue state.
+
+## Consequences
+
+Web stop preserves accepted queued intent and advances it automatically after truthful cancellation settlement. Queue rows may remain visible while uncooperative active work winds down, and external steering preserved by the same inbox option can enter the next admitted turn even though Web does not render steering in QueueDock. A future bulk-clear interaction requires an explicit product action rather than overloading stop.
@@ -0,0 +1,37 @@
+# Agent Note: Web 停止操作保留待处理 Queue
+
+Status: implemented
+
+[English](2026-07-31-web-stop-preserves-queue.md) | 中文
+
+## 问题
+
+Web 停止按钮调用 `session.cancel`，后者映射到广义 `agent.cancel({ kind: 'user' })`。在活动轮次期间，普通 composer 提交已经被接纳为可独立寻址的 Queue 入队项。用户只想停止当前生成时，广义取消却会丢弃所有入队项，混淆了轮次中断与 Queue 的显式删除操作。
+
+浏览器无法通过重发可见行修复这一损失。它不拥有这些行的实时 `InboxItemId`、唤醒策略或认领竞态；重发还可能重复 Host 已认领的工作。
+
+## 决策
+
+`session.cancel` 是 Web Host API 的活动轮次停止操作。它调用 `agent.cancel({ kind: 'user' }, { keepInbox: true })`，在协作式中止当前轮次的同时保留待处理 inbox 工作。底层选项会保留 queued 和 steering 入队项；Web Queue 投影继续只暴露 queued 入队项。
+
+AgentLoop 不会启动并发的替代轮次。它会关闭并 flush 被中断的轮次，达到取消的完全停稳，然后通过现有 FIFO 驱动器认领下一个可唤醒的 queued 入队项。该认领会发出 `agent/inbox/dequeue`，因此 Host 的权威 `session/queue` 快照会退役已认领行，并使剩余队尾保持可见。浏览器既不重发，也不提升任何行。忽略取消的工作会延迟这一交接，直到该工作结算。
+
+该映射只更改 Web 客户端使用的 Host `session.cancel` 端点。`Agent.cancel()` 默认契约仍为广义取消，ACP 和 TUI 保留既有取消策略，`AgentHandle.dispose()` 在拆卸期间仍会清除待处理工作。移除 Queue 行仍是用于丢弃单个待处理入队项的显式 Web 操作。
+
+## 考虑过的替代方案
+
+**停止按钮继续使用广义取消。** 之所以否决：停止一次生成不应销毁已独立排队的用户意图；Queue 已拥有显式删除操作。
+
+**取消后由浏览器重发下一行。** 之所以否决：Host 拥有入队项标识和认领顺序。客户端重新提交可能重复工作、重排 FIFO，或与权威出队产生竞态。
+
+**被取消工作达到完全停稳之前启动下一轮次。** 之所以否决：两个轮次会并发修改同一会话日志，并共享 Agent 拥有的资源。协作式取消会如实等待活动工作结算。
+
+**为广义取消与保留式取消添加协议选项。** 之所以否决：在 Web 产品提供独立的「停止并清空 Queue」交互之前，不需要此选项。现有停止按钮只有一项策略，而逐行删除已提供当前的丢弃控件。
+
+## 验证
+
+AgentLoop 覆盖会保持一个活动模型流，将两个可唤醒轮次排队，使用 `keepInbox` 取消，并固定验证先中止、后完成的轮次原因，FIFO 用户消息顺序，不存在 discard 事件，以及最终空闲状态。无密钥 Web 场景通过 HTTP／SSE 驱动已组装组合：它停止一个卡住的轮次，观察队尾保持可见时下一个 queued 入队项开始，再停止该轮次，并观察最后一个 queued 入队项完成。其可访问性快照固定了中间的 Queue 保留状态。
+
+## 后果
+
+Web 停止会保留已接纳的排队意图，并在取消如实结算后自动推进。不配合取消的活动工作收尾时，Queue 行可能仍保持可见；由同一 inbox 选项保留的外部 steering 可以进入下一个已接纳轮次，尽管 Web 不会在 QueueDock 中渲染 steering。未来的批量清空交互需要显式的产品操作，而不是过载停止。
@@ -2,5 +2,5 @@
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
 #   pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-07-29-addressable-queue-operations.md
-2026-07-29-addressable-queue-operations.md: 7a08b889c958e583dc430d33a1855fe3725f3d48
-2026-07-29-addressable-queue-operations.zh.md: 701b028c7494fd7cb608d05a5d170c9075b155d7
+2026-07-29-addressable-queue-operations.md: 57527730d0f43a3a6c7801806fb9cc136daa5f14
+2026-07-29-addressable-queue-operations.zh.md: 65617e492c16e41d19ba296c7d117f8a2771a67d
@@ -18,7 +18,7 @@ The Web queue rendered pending messages but could not edit or delete one row. `M

 **Queue addresses require a live Agent.** `session.updateQueue` queries only the mounted Agent registry and never resumes a cold session: an `InboxItemId` is process-local and cannot name work after restart or disposal. A missing Agent and a driver-claimed occurrence both return `queue-item-not-found`.

-**Web actions address Queue only.** The Host excludes pending steering from `session/queue`; steering retains its existing durable transcript path after consumption. QueueDock hides while empty, renders one pending occurrence directly, and defaults two or more occurrences to a collapsed `"<n> 条排队消息"` header that expands or collapses the complete list. The header exposes `aria-expanded` and `aria-controls`; the expanded list scrolls within a 180px height bound. An active edit or mutation keeps its rows visible, and emptying the queue restores the collapsed default for the next queue. Visible rows expose edit and delete, but no send-now control. The UI derives queue row and mutation types from the runtime `SessionFace` contract rather than importing the connection plugin, so plugin cooperation continues through services and snapshots. Edit is available only when all content blocks are text; the editor cannot silently drop non-text blocks. An editing row exposes only save and cancel, with Enter and Escape as their keyboard equivalents. Delete removes the exact occurrence.
+**Web actions address Queue only.** The Host excludes pending steering from `session/queue`; steering retains its existing durable transcript path after consumption. QueueDock hides while empty, renders one pending occurrence directly, and defaults two or more occurrences to a collapsed `"<n> 条排队消息"` header that expands or collapses the complete list. The header exposes `aria-expanded` and `aria-controls`; the expanded list scrolls within a 180px height bound. An active edit or mutation keeps its rows visible, and emptying the queue restores the collapsed default for the next queue. Visible rows expose edit and delete, but no send-now control. The UI derives queue row and mutation types from the runtime `SessionFace` contract rather than importing the connection plugin, so plugin cooperation continues through services and snapshots. Edit is available only when all content blocks are text; the editor cannot silently drop non-text blocks. An editing row exposes only save and cancel, with Enter and Escape as their keyboard equivalents. Delete removes the exact occurrence. The Web stop action preserves pending Queue work; AgentLoop claims the next waking occurrence only after the interrupted turn reaches quiescence, and its dequeue event retires that row without a browser resend.

 ## Alternatives considered

@@ -34,10 +34,10 @@ The Web queue rendered pending messages but could not edit or delete one row. `M

 ## Verification

-AgentLoop contract tests hold prompt admission while editing and removing exact queued occurrences, reject mutations of steering occurrences, and verify the resulting independent turn and terminal lifecycle events. Host schema and proxy tests cover queued-only authoritative snapshots, synchronous re-entrant mutation order, reconnect, cold-Agent rejection, typed not-found errors, and the RPC transport. Client runtime and QueueDock tests cover non-optimistic projection, single-row presentation, default multi-row collapse, interaction-forced visibility, reset after emptying, expansion, text-only editing, save and cancel affordances, removal, retirement races, and disabled mixed-content editing. Keyless browser scenarios capture the default collapsed header before expanding the queue and driving its exposed edit and delete actions through the built Web composition and real HTTP/SSE wire.
+AgentLoop contract tests hold prompt admission while editing and removing exact queued occurrences, reject mutations of steering occurrences, and verify the resulting independent turn and terminal lifecycle events. Host schema and proxy tests cover queued-only authoritative snapshots, synchronous re-entrant mutation order, reconnect, cold-Agent rejection, typed not-found errors, and the RPC transport. Client runtime and QueueDock tests cover non-optimistic projection, single-row presentation, default multi-row collapse, interaction-forced visibility, reset after emptying, expansion, text-only editing, save and cancel affordances, removal, retirement races, and disabled mixed-content editing. A keyless browser scenario captures the default collapsed header, drives edit and delete through the built Web composition and real HTTP/SSE wire, then stops consecutive active turns to prove the preserved FIFO advances without clearing its tail.

 ## Consequences

-Queued work gains precise row operations without becoming durable session history. Occurrence identity is a live process-local capability and disappears at claim, cancellation, disposal, or restart; reconnect recovers only queued items still held by the live Agent. Editing excludes mixed content until an editor can preserve every block, while pending steering remains outside this operation surface.
+Queued work gains precise row operations without becoming durable session history. Occurrence identity is a live process-local capability and disappears at claim, broad cancellation, disposal, or restart; the Web stop action preserves it until a later claim, while reconnect recovers only queued items still held by the live Agent. Editing excludes mixed content until an editor can preserve every block, while pending steering remains outside this operation surface.

 The protocol now carries full queue snapshots on each change. Queues are expected to remain short, so deterministic recovery and multi-client convergence are preferred over an incremental mutation protocol.
@@ -18,7 +18,7 @@ Web 队列能够渲染待处理消息，但无法编辑或删除其中某一行

 **Queue 寻址要求 Agent 存活。** `session.updateQueue` 只查询已挂载的 Agent 注册表，绝不恢复冷会话：`InboxItemId` 属于进程本地标识，无法在重启或资源释放后继续指向工作。Agent 缺失和单次入队项已被驱动器认领这两种情况都返回 `queue-item-not-found`。

-**Web 操作只面向 Queue。** Host 从 `session/queue` 中排除待处理 steering；steering 消费后仍沿用既有的持久 transcript（文本记录）路径。QueueDock 在队列为空时隐藏，只有一个待处理项时直接渲染该行，存在两个或更多待处理项时则默认收起为可展开或收起完整列表的 `"<n> 条排队消息"` 表头。表头暴露 `aria-expanded` 和 `aria-controls`；展开后的列表以 180px 为高度上限，并可滚动。存在进行中的编辑或变更时，列表行会保持可见；队列清空后，下一次出现队列时会恢复默认收起状态。可见行暴露编辑和删除操作，不提供立即发送控件。UI 从运行时 `SessionFace` 契约派生队列行与变更类型，而不是导入连接插件，因此插件仍通过服务和快照协作。仅当所有内容块都是文本时才提供编辑功能；编辑器不能静默丢弃非文本块。编辑中的行只展示保存和取消操作，对应的键盘操作分别是 Enter 和 Escape。删除会移除对应的精确入队项。
+**Web 操作只面向 Queue。** Host 从 `session/queue` 中排除待处理 steering；steering 消费后仍沿用既有的持久 transcript（文本记录）路径。QueueDock 在队列为空时隐藏，只有一个待处理项时直接渲染该行，存在两个或更多待处理项时则默认收起为可展开或收起完整列表的 `"<n> 条排队消息"` 表头。表头暴露 `aria-expanded` 和 `aria-controls`；展开后的列表以 180px 为高度上限，并可滚动。存在进行中的编辑或变更时，列表行会保持可见；队列清空后，下一次出现队列时会恢复默认收起状态。可见行暴露编辑和删除操作，不提供立即发送控件。UI 从运行时 `SessionFace` 契约派生队列行与变更类型，而不是导入连接插件，因此插件仍通过服务和快照协作。仅当所有内容块都是文本时才提供编辑功能；编辑器不能静默丢弃非文本块。编辑中的行只展示保存和取消操作，对应的键盘操作分别是 Enter 和 Escape。删除会移除对应的精确入队项。Web 停止操作会保留待处理 Queue 工作；只有在被中断轮次达到完全停稳后，AgentLoop 才认领下一个可唤醒入队项，其出队事件会退役该行，无需浏览器重发。

 ## 考虑过的替代方案

@@ -34,10 +34,10 @@ Web 队列能够渲染待处理消息，但无法编辑或删除其中某一行

 ## 验证

-AgentLoop 契约测试会在编辑和移除精确 queued 入队项时阻塞提示词接纳，拒绝对 steering 入队项的变更，并验证所得独立轮次及终态生命周期事件。Host schema 与代理测试覆盖仅含 queued 项的权威快照、同步可重入变更顺序、重连、拒绝冷 Agent、类型化 not-found 错误和 RPC 传输。客户端运行时和 QueueDock 测试覆盖非乐观投影、单行呈现、多行默认收起、交互期间强制保持可见、清空后重置、展开、仅文本编辑、保存与取消入口、移除、退役竞态，以及禁用混合内容编辑。无密钥浏览器场景会先捕获默认收起的表头，再展开队列，并通过构建后的 Web 组合和真实 HTTP／SSE 协议操作其公开的编辑和删除。
+AgentLoop 契约测试会在编辑和移除精确 queued 入队项时阻塞提示词接纳，拒绝对 steering 入队项的变更，并验证所得独立轮次及终态生命周期事件。Host schema 与代理测试覆盖仅含 queued 项的权威快照、同步可重入变更顺序、重连、拒绝冷 Agent、类型化 not-found 错误和 RPC 传输。客户端运行时和 QueueDock 测试覆盖非乐观投影、单行呈现、多行默认收起、交互期间强制保持可见、清空后重置、展开、仅文本编辑、保存与取消入口、移除、退役竞态，以及禁用混合内容编辑。一个无密钥浏览器场景会捕获默认收起的表头，通过构建后的 Web 组合和真实 HTTP／SSE 协议执行编辑和删除，随后连续停止活动轮次，证明保留的 FIFO 会继续推进且不清空队尾。

 ## 后果

-queued 工作获得精确的行操作，但不会因此成为持久会话历史。单次入队标识是进程本地的实时寻址凭据，会在认领、取消、dispose 或重启时消失；重连只能恢复仍由活跃 Agent 持有的 queued 项。编辑会排除混合内容，直至编辑器能够保留每个块；待处理 steering 则不属于此操作接口。
+queued 工作获得精确的行操作，但不会因此成为持久会话历史。单次入队标识是进程本地的实时寻址凭据，会在认领、广义取消、dispose 或重启时消失；Web 停止操作会将其保留到后续认领，而重连只能恢复仍由活跃 Agent 持有的 queued 项。编辑会排除混合内容，直至编辑器能够保留每个块；待处理 steering 则不属于此操作接口。

 现在，协议会在每次变更时携带完整队列快照。队列预期保持较短，因此系统优先选择确定性恢复和多客户端收敛，而非增量变更协议。
@@ -1,6 +1,6 @@
 # Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
-#   pnpm run verify-translation-pairing --write
-2026-06-20-public-agent-stop-surface.md: e22c4389df18f3c9ca96763fc097eabefcc5b761
-2026-06-20-public-agent-stop-surface.zh.md: e2647b498a8c906579b4fd2b50f94d1c326fe784
+#   pnpm run verify-translation-pairing --write .agents/notes/implemented/simplification/2026-06-20-public-agent-stop-surface.md
+2026-06-20-public-agent-stop-surface.md: 7e8f6f691c999fd78c9b4133eaeac40f2d1c1ba9
+2026-06-20-public-agent-stop-surface.zh.md: 041d666eb817450add2d7a1746f7f5b81223568a
@@ -8,15 +8,15 @@ English | [中文](2026-06-20-public-agent-stop-surface.zh.md)

 ## Problem

-The public `Agent` handle exposed two overlapping ways to stop in-flight work: step-only `abort()` and queue-aware `cancel()`. The former preserved queued input while the latter clears queued and steering work and aborts the active turn. In production, ACP uses `cancel()` for `session/cancel`, while lifecycle owners tear down agents through `AgentHandle.dispose()`. No production caller needs a bare step-only abort.
+The public `Agent` handle exposed two overlapping ways to stop in-flight work: step-only `abort()` and queue-aware `cancel()`. The former preserved queued input while the latter originally only exposed its broad default, which clears queued and steering work while aborting the active turn. `cancel(cause, { keepInbox: true })` now covers the production Web stop policy without exposing the private turn holder; ACP retains broad cancellation, while lifecycle owners tear down agents through `AgentHandle.dispose()`. No production caller needs a bare step-only abort.

-The behavioral distinction is real, but no shipping code needs the narrower operation. AgentLoop instead owns one private cancellation holder for the whole turn. `cancel(cause?)` carries a typed `user` or `parent` cause, defaults to `user`, and drops pending input; disposal remains a separate lifecycle interruption. The complete ownership and propagation contract lives in the [explicit turn cancellation RFC](../architecture/2026-07-16-explicit-turn-cancellation.md).
+The behavioral distinction is real, but no shipping code needs a separate narrower verb. AgentLoop owns one private cancellation holder for the whole turn. `cancel(cause, options?)` carries an explicit typed `user` or `parent` cause; its broad default drops pending input, while `keepInbox` preserves pending work for later turns. Disposal remains a separate lifecycle interruption. The complete ownership and propagation contract lives in the [explicit turn cancellation RFC](../architecture/2026-07-16-explicit-turn-cancellation.md).

-The extra surface area made the loop carry a public verb that is mostly a teardown internal: `abort()` had to be documented as distinct from queue-aware cancellation even though a UI cancellation almost always wants the broader operation.
+The extra surface area made the loop carry a public verb that was mostly a teardown internal. An options-bearing `cancel()` expresses caller policy without exposing a second holder-shaped operation.

 ## Decision

-`cancel()` is the only public *stop* primitive on `Agent`. Lifecycle owners use `AgentHandle.dispose()` to stop and unregister an agent; non-owners use `cancel()` to abandon current and queued work. The implementation keeps a private turn cancellation holder, but it is not part of the plugin-facing `Agent` contract.
+`cancel()` is the only public *stop* primitive on `Agent`. Lifecycle owners use `AgentHandle.dispose()` to stop and unregister an agent; non-owners use broad `cancel()` to abandon current and queued work or `keepInbox` to abort the active turn while retaining pending work. The implementation keeps a private turn cancellation holder, but it is not part of the plugin-facing `Agent` contract. The [Web stop decision](../bug-fix/2026-07-31-web-stop-preserves-queue.md) is the production `keepInbox` consumer.

 `whenIdle()` is **retained** as the public quiescence-observation primitive (resolve once the agent settles out of `running`, resolve immediately when already idle, await the loop exit when disposed). It is not a stop verb; it is how a non-owner observes the stop *completing* without disposing the agent. Its live consumers are ACP and agent tests that await settlement through this public seam (`packages/acp/acp/tests`, `packages/core/agent-loop/tests`); the production ACP bridge owns its agents and tears them down through `AgentHandle.dispose()`, so `packages/acp/acp/src` itself has no `whenIdle()` call.

@@ -28,11 +28,11 @@ Public `abort()` is absent, and the disposer remains async and waits for the loo

 ## Verification

-`Agent` exposes no public `abort()` while `cancel()`, `whenIdle()`, and `steer()` remain; ACP cancellation calls `cancel()`; teardown awaits quiescence through handle disposal, with `whenIdle()` resolving on quiescence for non-owner observers; and the suites cover cancellation and disposal as the two supported stop paths.
+`Agent` exposes no public `abort()` while `cancel()`, `whenIdle()`, and `steer()` remain; ACP cancellation calls broad `cancel()`, Web stop calls `cancel(..., { keepInbox: true })`, and teardown awaits quiescence through handle disposal. `whenIdle()` resolves on quiescence for non-owner observers, and the suites cover cancellation and disposal as the two supported stop paths.

 ## Consequences

-A future plugin cannot abort only the current model/tool step while preserving queued prompts through the public interface. If that use case becomes real, it should return with a named consumer and a narrower contract. Today it is latent generality that keeps a private loop mechanic public.
+A plugin can abort the active turn while preserving queued prompts through `keepInbox`, but it cannot abort only one model/tool step while leaving that turn running. A step-only use case would need a named consumer and a narrower contract; exposing the private loop mechanic remains unjustified.

 ## Related

@@ -8,15 +8,15 @@ Status: implemented

 ## 问题

-公共 `Agent` handle 暴露了两种相互重叠的在途工作停止方式：仅针对步骤的 `abort()` 和感知队列的 `cancel()`。前者保留已排队输入，后者则清除已排队和 steering（中途引导）工作，并中止活动轮次。在生产中，ACP（Agent Client Protocol）对 `session/cancel` 使用 `cancel()`，生命周期拥有者则通过 `AgentHandle.dispose()` 拆除 agent（智能体）。没有生产调用方需要一个裸的、仅针对步骤的 abort。
+公共 `Agent` handle 暴露了两种相互重叠的在途工作停止方式：仅针对步骤的 `abort()` 和感知队列的 `cancel()`。前者保留已排队输入，后者原本只暴露广义默认行为，该行为会清除已排队和 steering（中途引导）工作，同时中止活动轮次。`cancel(cause, { keepInbox: true })` 现在无需暴露私有轮次 holder 即可覆盖生产环境的 Web 停止策略；ACP 保留广义取消，生命周期拥有者则通过 `AgentHandle.dispose()` 拆除 agent（智能体）。没有生产调用方需要一个裸的、仅针对步骤的 abort。

-行为差异确实存在，但已发布代码不需要较窄的操作。AgentLoop 改为为整个轮次拥有一个私有取消 holder。`cancel(cause?)` 携带类型化的 `user` 或 `parent` 原因，默认为 `user`，并丢弃待处理输入；释放仍是单独的生命周期中断。完整的归属与传播契约位于[显式轮次取消 Agent Note（agent 决策记录）](../architecture/2026-07-16-explicit-turn-cancellation.md)。
+行为差异确实存在，但已发布代码不需要独立的更窄动词。AgentLoop 为整个轮次拥有一个私有取消 holder。`cancel(cause, options?)` 携带显式且类型化的 `user` 或 `parent` 原因；其广义默认行为丢弃待处理输入，`keepInbox` 则为后续轮次保留待处理工作。资源释放仍是单独的生命周期中断。完整的归属与传播契约位于[显式轮次取消 Agent Note（agent 决策记录）](../architecture/2026-07-16-explicit-turn-cancellation.md)。

-多余的公开接口使得循环不得不承载一个本质上属于内部拆卸的公开动词：`abort()` 必须被文档描述为有别于队列感知的取消，尽管 UI 取消几乎总是需要更广泛的操作。
+多余的公开接口使循环承载了一个本质上属于内部拆卸的公开动词。带选项的 `cancel()` 可以表达调用方策略，而无需暴露第二个 holder 形态的操作。

 ## 决策

-`cancel()` 是 `Agent` 上唯一的公共*停止*原语。生命周期拥有者使用 `AgentHandle.dispose()` 停止并注销 agent；非拥有者使用 `cancel()` 放弃当前和已排队工作。实现保留一个私有轮次取消 holder，但它不属于面向插件的 `Agent` 契约。
+`cancel()` 是 `Agent` 上唯一的公共*停止*原语。生命周期拥有者使用 `AgentHandle.dispose()` 停止并注销 agent；非拥有者使用广义 `cancel()` 放弃当前和已排队工作，或使用 `keepInbox` 中止活动轮次并保留待处理工作。实现保留一个私有轮次取消 holder，但它不属于面向插件的 `Agent` 契约。[Web 停止决策](../bug-fix/2026-07-31-web-stop-preserves-queue.md)是生产环境中的 `keepInbox` 消费方。

 `whenIdle()` **保留**为公开的完全停稳观测原语（agent 从 `running` 状态稳定后 resolve，已处于 idle 时立即 resolve，dispose 后等待循环退出）。它不是停止动词；它是非所有者在不 dispose agent 的前提下观测停止*完成*的方式。它的活跃消费方是 ACP 和通过此公开 seam 等待结算的 agent 测试（`packages/acp/acp/tests`、`packages/core/agent-loop/tests`）；生产环境的 ACP 桥接层拥有其 agent 并通过 `AgentHandle.dispose()` 销毁它们，因此 `packages/acp/acp/src` 本身没有 `whenIdle()` 调用。

@@ -28,11 +28,11 @@ Status: implemented

 ## 验证

-`Agent` 不再暴露公开的 `abort()`，而 `cancel()`、`whenIdle()` 和 `steer()` 保留；ACP 取消调用 `cancel()`；拆卸通过 handle disposal 等待完全停稳，`whenIdle()` 在完全停稳时为非所有者观测者 resolve；测试套件覆盖取消和 disposal 作为两条受支持的停止路径。
+`Agent` 不再暴露公开的 `abort()`，而 `cancel()`、`whenIdle()` 和 `steer()` 保留；ACP 取消调用广义 `cancel()`，Web 停止调用 `cancel(..., { keepInbox: true })`，拆卸则通过 handle disposal 等待完全停稳。`whenIdle()` 在完全停稳时为非所有者观测者 resolve；测试套件覆盖取消和资源释放这两条受支持的停止路径。

 ## 后果

-未来的插件无法通过公开接口仅中止当前模型/工具步骤而保留队列中的提示词。如果该用例变为现实需求，它应当带着一个具名消费方和更窄的契约回归。目前它是将私有循环机制保持公开的潜在泛化。
+插件可以通过 `keepInbox` 在保留已排队提示词的同时中止活动轮次，但不能只中止某一个模型／工具步骤而让该轮次继续运行。仅步骤用例需要具名消费方和更窄契约；暴露私有循环机制仍缺乏正当理由。

 ## 相关

@@ -1,16 +1,16 @@
 // Keyless browser coverage for pending queue actions through the shipped Web
-// composition and real HTTP/SSE wire. A replay override parks the active turn
-// so two ordinary follow-ups remain addressable while the page edits one and
-// removes one. The queue uses an existing recorded model
-// call; this scenario owns only the user-visible mid-turn golden.
+// composition and real HTTP/SSE wire. Replay overrides park consecutive turns
+// so the page can edit and remove exact occurrences, then stop the active turn
+// while proving the preserved Queue advances in FIFO order.
 import { existsSync } from 'node:fs'
-import { mkdtemp, rm, writeFile } from 'node:fs/promises'
+import { mkdtemp, readFile, rm, writeFile } from 'node:fs/promises'
 import { tmpdir } from 'node:os'
 import { fileURLToPath } from 'node:url'
 import { join } from 'node:path'
 import type { Browser, Page } from 'playwright'
 import { chromium } from 'playwright'
 import { afterEach, describe, expect, it, onTestFailed } from 'vitest'
+import { deriveReplayScript, parseSessionLog, type ReplayEntry } from '@deepseek-ai/dsh-llm-replay'
 import type { SessionEvent } from '@deepseek-ai/dsh-session'
 import {
  assertFixtureInventory, captureStableAria, compareOrRefreshGolden,
@@ -22,6 +22,7 @@ const SNAPSHOT_DIR = fileURLToPath(new URL('./snapshots/queue-actions', import.m
 const FIXTURE = fileURLToPath(new URL('./snapshots/live-interactions/session.jsonl', import.meta.url))
 const COLLAPSED_EXPECTED = join(SNAPSHOT_DIR, 'collapsed.expected.md')
 const EDITING_EXPECTED = join(SNAPSHOT_DIR, 'editing.expected.md')
+const PRESERVED_EXPECTED = join(SNAPSHOT_DIR, 'preserved.expected.md')
 const UI_EXPECTED = join(SNAPSHOT_DIR, 'ui.expected.md')
 const MODE = webSnapshotMode()

@@ -29,6 +30,12 @@ const ACTIVE_PROMPT = 'Reply with a one-sentence description of event sourcing,
 const REMOVE = 'Queue item to remove'
 const EDIT = 'Queue item to edit'
 const EDITED = 'Edited queue item'
+const TAIL = 'Queue item preserved after stop'
+
+/** Durable turn-end classifications observed by the scenario. */
+function turnEndReasons(events: readonly SessionEvent[]): string[] {
+  return events.flatMap(event => event.type === 'turn/end' ? [event.data.reason.kind] : [])
+}

 describe('web e2e: queue row actions', () => {
  let scaffold: WebScaffold | undefined
@@ -52,13 +59,19 @@ describe('web e2e: queue row actions', () => {
    if (failures.length > 1) throw new AggregateError(failures, 'queue-actions teardown failed')
  })

-  it.skipIf(MODE === 'record')('edits and removes exact pending occurrences', async () => {
+  it.skipIf(MODE === 'record')('edits and removes exact occurrences and preserves Queue across stop', async () => {
    overrideDir = await mkdtemp(join(tmpdir(), 'dsh-web-queue-actions-'))
    const readyFile = join(overrideDir, '.hang-ready')
+    const nextReadyFile = join(overrideDir, '.next-hang-ready')
    const overridePath = join(overrideDir, 'replay.override.json')
-    await writeFile(overridePath, JSON.stringify({
-      patches: [{ at: 0, entry: { kind: 'hang', readyFile } }],
-    }))
+    const recorded = deriveReplayScript(parseSessionLog(await readFile(FIXTURE, 'utf8')))
+    expect(recorded).toHaveLength(1)
+    const replay: ReplayEntry[] = [
+      { kind: 'hang', readyFile },
+      { kind: 'hang', readyFile: nextReadyFile },
+      recorded[0]!,
+    ]
+    await writeFile(overridePath, JSON.stringify(replay))

    const sessionEvents: SessionEvent[] = []
    scaffold = await launchWebScaffold({ replayFixture: FIXTURE, replayOverride: overridePath })
@@ -135,17 +148,34 @@ describe('web e2e: queue row actions', () => {
    expect(tripwire.pageErrors).toEqual([])
    expect(tripwire.warnings).toEqual([])

-    const editedRow = page.getByText(EDITED, { exact: true }).locator('..')
-    await editedRow.getByRole('button', { name: 'Remove queued message' }).click()
-    await expect.poll(() => page.getByText(EDITED, { exact: true }).count()).toBe(0)
+    await input.fill(TAIL)
+    await input.press('Enter')
+    await expect.poll(
+      () => page.getByRole('button', { name: 'Remove queued message' }).count(),
+      { timeout: 10_000 },
+    ).toBe(2)
+
+    await page.getByRole('button', { name: 'Stop generating' }).click()
+    await expect.poll(() => existsSync(nextReadyFile), { timeout: 15_000 }).toBe(true)
+    await page.getByText(TAIL, { exact: true }).waitFor()
+    await expect.poll(() => page.getByRole('button', { name: 'Remove queued message' }).count())
+      .toBe(1)
+
+    const preservedSnapshot = await captureStableAria(page, '[class*="centerCol"]', scaffold.workspaceCwd)
+    await compareOrRefreshGolden(PRESERVED_EXPECTED, preservedSnapshot, MODE)
+
    await page.getByRole('button', { name: 'Stop generating' }).click()
    await settled
+    expect(turnEndReasons(sessionEvents)).toEqual(['aborted', 'aborted', 'completed'])
+    expect(sessionEvents.filter(event => event.type === 'user/message' && event.data.source.kind === 'user'))
+      .toHaveLength(3)
+    await expect.poll(() => page.locator('[data-queue-dock]').count()).toBe(0)
  }, 120_000)

  it.skipIf(MODE === 'record')('keeps its snapshot inventory closed', async () => {
    await assertFixtureInventory(
      SNAPSHOT_DIR,
-      ['collapsed.expected.md', 'editing.expected.md', 'ui.expected.md'],
+      ['collapsed.expected.md', 'editing.expected.md', 'preserved.expected.md', 'ui.expected.md'],
    )
  })
 })
@@ -0,0 +1,44 @@
+- banner:
+  - navigation "Session hierarchy":
+    - button "Reply with a one-sentence description" [disabled]
+  - tablist:
+    - tab "Chat" [selected]
+    - tab "Trajectory"
+- text: Reply with a one-sentence description of event sourcing, then stop. {{clock}}
+- button "Copy":
+  - img
+- button "Branch into a new conversation":
+  - img
+- button "Context injection":
+  - img
+  - img
+  - text: Context injection
+- paragraph: partial
+- text: Stopped
+- button "Copy":
+  - img
+- button "Branch into a new conversation":
+  - img
+- text: {{clock}} Edited queue item {{clock}}
+- button "Copy":
+  - img
+- button "Branch into a new conversation":
+  - img
+- paragraph: partial
+- status: Deep diving...
+- list:
+  - listitem:
+    - text: Queue item preserved after stop
+    - button "Edit queued message":
+      - img
+    - button "Remove queued message":
+      - img
+- textbox "Message the agent"
+- button "Commands":
+  - img
+- 'button "Access mode, current: Full access"': Full access
+- button "Select model, current DeepSeek-V4-Flash":
+  - text: DeepSeek-V4-Flash
+  - img
+- button "Stop generating"
+- text: 1 turns · 1 steps Input 0 tok · Output 0 tok
@@ -46,7 +46,8 @@ export interface ISession {
   */
  updateQueue(itemId: InboxItemId, action: QueueAction): Promise<RpcResult<{ accepted: true }>>
  /**
-   * Cancel the running turn.
+   * Cancel the running turn. Pending queued work remains and resumes in FIFO
+   * order after the Host reaches cancellation quiescence.
   * @returns acceptance, or the business error.
   */
  cancel(): Promise<RpcResult<{ accepted: true }>>
@@ -248,7 +248,8 @@ export class Session implements SessionFace {
  }

  /**
-   * Stop: contract session.cancel 1:1; failures land in promptError (same error-strip display slot).
+   * Stop the active turn while the Host preserves pending inbox work; failures
+   * land in promptError (same error-strip display slot).
   * @returns the cancel result.
   */
  async cancel(): Promise<RpcResult<{ accepted: true }>> {
@@ -38,7 +38,7 @@ export interface IConversation {
   */
  updateQueue(itemId: QueueItemId, action: QueueAction): Promise<void>
  /**
-   * Cancel the scoped session's in-flight turn.
+   * Cancel the scoped session's in-flight turn while preserving its pending Queue.
   * @returns completion; failures reject as in send.
   */
  cancel(): Promise<void>
@@ -86,7 +86,7 @@ export class ConversationService extends Service implements IConversation {
    }
  }

-  /** Cancel the scoped session's in-flight turn (failures land in promptError and reject, as in send). */
+  /** Cancel the scoped session's in-flight turn while preserving Queue (failures land in promptError and reject, as in send). */
  async cancel(): Promise<void> {
    const session = this.scopedSession('cancel')
    const result = await session.cancel()
@@ -1,9 +1,9 @@
 import { createUserMessage } from '@deepseek-ai/dsh-llm'
 /**
- * Tests for the queue-aware `Agent.cancel()` primitive. `cancel()` is the broad verb — it
- * clears queued + steering work, aborts the active turn, and drops work not yet claimed by the
- * driver without leaking cancellation into a replacement prompt. The suite covers every landing
- * window plus signal reset and `whenIdle()` quiescence.
+ * Tests for the queue-aware `Agent.cancel()` primitive. The default clears
+ * queued and steering work, while `keepInbox` preserves pending input and
+ * resumes waking turns after the active turn reaches quiescence. The suite
+ * covers every landing window plus signal reset and `whenIdle()` quiescence.
 * @module dsh-agent-loop/tests/cancel
 */

@@ -302,6 +302,41 @@ describe('Agent.cancel()', () => {
    expect(adapter.requests).toHaveLength(1)
  })

+  it('cancel({ keepInbox: true }) aborts the active turn and drains the queued tail in FIFO order', async () => {
+    const adapter = new MockAdapter([
+      'hang',
+      textResponse('second reply'),
+      textResponse('third reply'),
+    ])
+    const ctx = await harness(adapter)
+    const agent = ctx.agentLoop.create(SessionId('keep-inbox-running'), { provider: 'mock', model: 'mock' })
+    const reasons: TurnEndReason[] = []
+    const discards: unknown[] = []
+    ctx.on('session/event', (session, event) => {
+      if (session === agent.session && event.type === 'turn/end') reasons.push(event.data.reason)
+    })
+    ctx.on('agent/inbox/discard', (subject, items) => {
+      if (subject === agent) discards.push(items)
+    })
+
+    send(agent, 'active')
+    await new Promise(resolve => setTimeout(resolve, 30))
+    send(agent, 'queued second')
+    send(agent, 'queued third')
+    const idle = agent.whenIdle()
+    agent.cancel({ kind: 'user' }, { keepInbox: true })
+    await idle
+
+    expect(discards).toEqual([])
+    expect(userTexts(agent)).toEqual(['active', 'queued second', 'queued third'])
+    expect(reasons).toEqual([
+      { kind: 'aborted' },
+      { kind: 'completed' },
+      { kind: 'completed' },
+    ])
+    expect(adapter.requests).toHaveLength(3)
+  })
+
  it('cancel from an assistant/message observer skips execution but balances replay', async () => {
    const adapter = new MockAdapter([
      toolCallResponse('c1', 'danger', {}),
@@ -2,5 +2,5 @@
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
 #   pnpm run verify-translation-pairing --write packages/host/apiproxy/README.md
-README.md: 5af2739ee419d32c93578b37bebc6808431fd003
-README.zh.md: c2729c379bd3c132c94d96134c04a0976a3aca9b
+README.md: 3c5a83a468b0cf9e596b8b13fafe40c409576fc5
+README.zh.md: f8533564575bf6b716f3fa7241ce47b8d4dd435f
@@ -20,7 +20,7 @@ Session titles ride the generic projection pair like every other domain — the

 Session model routing is a session-domain contract. `session.models` returns the selected provider/model/reasoning target with provider-grouped advisory models, exact-route reasoning metadata, and provider-local lookup failures. `session.selectModel` validates the optional adapter-owned reasoning effort and replaces the complete target selected for the next prompt-assembly boundary. Catalog membership is not validation: an adapter may resolve an unlisted model, while an unavailable route or unsupported effort returns `model-unavailable`.

-Pending queued input is a live control-plane contract, not session history. The gateway mirrors queued `InboxItem` occurrences from `agent/inbox/*` and broadcasts authoritative `session/queue` snapshots on every queued change and reconnect; pending steering stays outside this Web projection. `session.updateQueue` addresses one `InboxItemId`: edit replaces pending content and remove discards it. A driver claim wins races by retiring the address before admission; a later operation returns `queue-item-not-found`. The operation queries only an attached Agent and never resumes a cold session because process-local inbox identities do not survive restart or disposal. The client never infers retirement from turn or status events.
+Pending queued input is a live control-plane contract, not session history. The gateway mirrors queued `InboxItem` occurrences from `agent/inbox/*` and broadcasts authoritative `session/queue` snapshots on every queued change and reconnect; pending steering stays outside this Web projection. `session.updateQueue` addresses one `InboxItemId`: edit replaces pending content and remove discards it. `session.cancel` aborts only the active turn and preserves pending inbox work; after cancellation reaches quiescence and the closing turn flushes, AgentLoop claims the next waking occurrence in FIFO order. The browser never resends or promotes that occurrence. A driver claim wins races by retiring the address before admission; a later operation returns `queue-item-not-found`. Queue operations query only an attached Agent and never resume a cold session because process-local inbox identities do not survive restart or disposal. The client never infers retirement from turn or status events.

 Workspace and Session lists are separate reconnect baselines. `workspace.create({ name })` creates a uniquely titled directory under the configured root, while `workspace.create({ path })` adopts an existing canonical directory and permits basename-derived titles to repeat. `workspace.delete` removes only the Workspace registration, `session.create` accepts an optional preallocated Session id, and `host/workspace-changed`, `host/workspace-removed`, plus `host/session-added` carry committed increments in either arrival order. `workspace.archiveSession` adds one session to the registry-global archive set and answers the full updated set; `workspace.list` carries that set as the reconnect baseline and `host/archived-sessions-changed` pushes the full snapshot after every durable change. Archiving hides the session from grouping surfaces without touching its log or its workspace account; a session neither live nor persisted fails with `session-not-found`. Registration deletion preserves the directory and session logs; its Sessions remain in `session.list` and become Ungrouped. `SessionSummary.blank` and the `host/session-added` frame carry the derived zero-events bit: clients hide blank sessions and reuse them per workspace, flip blank on the first `host/session-status(running:true)`, and treat `session.list` as the reconnect authority; cold summaries are never blank because lazy persistence keeps never-appended sessions out of `list()`.

@@ -20,7 +20,7 @@

 会话模型路由属于会话领域契约。`session.models` 返回选中的提供方／模型／推理目标，以及按提供方分组的建议性模型、精确路由推理元数据和逐提供方查询失败记录。`session.selectModel` 校验由适配器持有的可选推理强度，并替换将在下一提示词组装边界使用的完整目标。目录成员关系不构成校验：适配器可以解析未列出的模型，而不可用路由或不受支持的推理强度会返回 `model-unavailable`。

-待处理的 queued 输入属于实时控制平面契约，而非会话历史。网关镜像来自 `agent/inbox/*` 的 queued `InboxItem` 入队项，并在每次 queued 变更和重连时广播权威的 `session/queue` 快照；待处理 steering（中途引导）不进入此 Web 投影。`session.updateQueue` 通过 `InboxItemId` 寻址单个项：编辑会替换待处理内容，移除会将其丢弃。驱动器在接纳前退役寻址标识，因此认领会赢得竞态；之后的操作返回 `queue-item-not-found`。该操作只查询当前已挂载的 Agent，绝不恢复冷会话，因为进程本地 inbox 标识无法在重启或资源释放后存活。客户端绝不根据轮次或状态事件推断项已退役。
+待处理的 queued 输入属于实时控制平面契约，而非会话历史。网关镜像来自 `agent/inbox/*` 的 queued `InboxItem` 入队项，并在每次 queued 变更和重连时广播权威的 `session/queue` 快照；待处理 steering（中途引导）不进入此 Web 投影。`session.updateQueue` 通过 `InboxItemId` 寻址单个项：编辑会替换待处理内容，移除会将其丢弃。`session.cancel` 仅中止活动轮次，并保留待处理 inbox 工作；取消达到完全停稳且结束中的轮次完成 flush 后，AgentLoop 按 FIFO 顺序认领下一个可唤醒入队项。浏览器绝不重发或提升该入队项。驱动器在接纳前退役寻址标识，因此认领会赢得竞态；之后的操作返回 `queue-item-not-found`。队列操作只查询当前已挂载的 Agent，绝不恢复冷会话，因为进程本地 inbox 标识无法在重启或资源释放后存活。客户端绝不根据轮次或状态事件推断项已退役。

 Workspace 列表与 Session 列表是相互独立的重连基线。`workspace.create({ name })` 会在配置根目录下创建显示标题唯一的目录，而 `workspace.create({ path })` 会接纳已有的规范目录，并允许由 basename 派生的标题重复。`workspace.delete` 只移除 Workspace 注册记录，`session.create` 接受可选的预分配 Session id，`host/workspace-changed`、`host/workspace-removed` 与 `host/session-added` 则以任意到达顺序携带已提交的增量。`workspace.archiveSession` 向注册表级全局归档集合添加一个会话，并应答完整的更新后集合；`workspace.list` 携带该集合作为重连基线，`host/archived-sessions-changed` 在每次持久变更后推送完整快照。归档只把会话从各分组视图中隐藏，不触碰其日志和 workspace 记账；既非实时也未持久化的会话以 `session-not-found` 失败。删除注册记录会保留目录和会话日志；相关 Session 仍留在 `session.list` 中，并进入 Ungrouped。`SessionSummary.blank` 与 `host/session-added` 帧携带派生的零事件位：客户端隐藏空白会话并按 workspace 复用它们，在首个 `host/session-status(running:true)` 时翻转 blank，并以 `session.list` 作为重连权威；冷会话摘要永远不是空白：惰性持久化让从未追加过事件的会话根本不出现在 `list()` 中。

@@ -1587,7 +1587,7 @@ export function createApiProxy(ctx: Context, defaults: ApiProxyDefaults): ApiPro
            details: { sessionId },
          }))
        }
-        agent.cancel({ kind: 'user' })
+        agent.cancel({ kind: 'user' }, { keepInbox: true })
        return Promise.resolve(ok(request, { accepted: true as const }))
      },
    },
@@ -279,7 +279,7 @@ export interface SessionsApi {
  updateQueue(request: RpcRequest<{ sessionId: SessionId; itemId: InboxItemId; action: QueueAction }>):
  Promise<RpcResponse<{ accepted: true }>>

-  /** Stops: clears both FIFOs + aborts the current step (1:1 with agent.cancel). */
+  /** Stops the active turn, preserving pending inbox work that resumes in FIFO order after cancellation settles. */
  cancel(request: RpcRequest<{ sessionId: SessionId }>): Promise<RpcResponse<{ accepted: true }>>

 }